Artificial intelligence, journalism and the uncertain future of the public sphere

NYT Publisher A.G. Sulzberger
Damon Winter/The New York Times via AP
The era of artificial intelligence announced its arrival less than four years ago with the public launch of ChatGPT. In just a few months, OpenAI's chatbot has amassed 100 million users, making it the fastest-growing consumer product in history. Today, it is just one of several increasingly powerful AI systems, alongside those developed by Anthropic, Google, Meta, Microsoft and X. There is little doubt that generative artificial intelligence represents the next big technological revolution - and it brings with it a dizzying array of important questions. Will AI drive a leap in productivity? Are you going to eliminate entire categories of jobs? Will it unlock extraordinary medical advances? Or facilitate biological attacks? Is it possible to fully understand the actions of AI models and agents? Is it possible to control them?
I'm here today to talk about issues that are, I admit, a little more restricted. But they matter a lot to me, to you and to society.
How will AI change journalism? How will these changes affect the information ecosystem that serves as the public sphere for engaged citizens around the world? And what can the people in this room do to ensure the future of fact-based journalism and first-hand reporting - essential to the health of our democracies?
Early signs give us reason to worry
The companies leading AI, already among the richest and most powerful in human history, are consolidating disproportionate control over our data and our attention. At the same time, they fail to assume a fundamental responsibility that comes with that power: ensuring that the public has access to trustworthy news and information.
This hijacking of the public sphere is enabled by the original sin that powers their AI products - a brazen theft of intellectual property on an unprecedented scale. Tech giants scrape news sites without permission and without compensation. They repackage stolen material as their own, diverting audiences and revenue that should go to the news organizations that created this work. And this doesn't happen just once, during the training process, but countless times, every day.
So I fear we are quickly moving toward a future with fewer and fewer journalists able to do the expensive and difficult work of original reporting - going places, talking to people, seeking information, covering relevant topics and events, offering context and analysis, investigating the powerful. A future in which an essential source of a healthy society and stable democracy - the truth, understanding and accountability provided by original journalism - continues to be depleted.
This potential harm goes far beyond journalism. AI companies have plundered civilization's entire body of original works - an act that also threatens the future of books, films, music, scientific research, and a host of other fields. In the United States, these industries represent not only the heart of the country's cultural and intellectual life, but also a pillar of its economy and one of its most influential exports. Globally, creative professions employ more than 50 million people and generate about $12 trillion in economic value per year.
The people gathered here today lead news organizations from more than 60 countries. This means they have already experienced a series of pressures that have plagued journalism around the world - from falling revenues to technological intermediation and growing attacks on press freedom. But in the face of AI, we need to do more. Our profession has been too silent, too passive and too fragmented in the face of the abuses of the companies leading this revolution.
We cannot allow AI enthusiasts to dominate the public conversation without standing up for a sustainable future for original journalism.

We cannot watch as AI companies try to permanently dismantle the rights that give us control over the work we create. We cannot stand by while this work is used to build substitute products that undermine our ability to gain the audience and revenue we need to continue doing journalism.
Some technology industry leaders will portray my comments today as anti-AI. As a defense of the status quo. Like another rigid institution reacting angrily to the innovators who drive progress. And to be fair to our Silicon Valley colleagues, there is a tradition of established incumbents - say, a 175-year-old newspaper - complaining about new technologies and the disruptors behind them.
So it's worth saying clearly: The organization I lead, The New York Times, has a long history of embracing technology to advance the mission of independent journalism. We have a history of respectful partnerships with technology companies to bring this journalism to new readers, in new ways. Facing disruption with curiosity, openness, and adaptability helped us get through the collapse of our print business and emerge stronger on the other side. Today, my colleagues use AI technology - responsibly, ethically and with humans making the decisions - to improve how we report, edit, distribute and monetize our journalism. Keeping powerful new technology at bay is a recipe for failure.
And I fully believe that AI has the power to do a lot of good in the world. I'm not calling AI - nor the tech giants that control this technology - inherently bad or evil. I'm warning that AI companies are making choices that violate established laws, threaten the viability of creative work, and appear designed to cause unnecessary and serious harm.
News organizations should want the benefits that AI can bring. But technology companies should also want to support the healthy, sustainable flow of information, ideas and creativity that fuels AI itself - to ensure that their actions don't lead us to a tragedy of the civic commons.
The four ingredients of AI
AI models are made with four basic ingredients.
The first is talent - the people who develop the algorithms. The second is what technology companies call "compute": the infrastructure behind AI, like chips and data centers. The third is energy - the electricity needed to power these resource-consuming products. The fourth is what technology companies call "data." The word itself seems almost designed to make creative and expressive work sound trivial, like an abundant commodity. But "data" is often used, among other things, as a synonym for books, films, music and journalism - what could more accurately be described as "copyrighted content". No tech CEO would dare suggest that the most talented engineers work for free. On the contrary, they regularly offer compensation packages that run into the tens or even hundreds of millions of dollars. Nor would they consider stealing chips from an Nvidia factory or making an illegal connection to a power line. Investors consider the potential financial rewards of AI to be so great that they are accepting losses in the hundreds of billions of dollars to build data centers and power plants.
In contrast, AI companies take "data" without consent or compensation. Justifications for theft change all the time. They say innovation requires this. They insist that they are just using facts, which no one can possess. They complain that agreements take too long and cost too much.

They claim that the doctrine of "fair use" allows them to take content for free anyway. Sometimes they even invoke national security - warning that if AI companies are forced to pay, the United States will lose the technology race to China.
None of these arguments stands up to scrutiny. A chatbot can only reproduce "facts" because it has illegally copied entire news articles, which allows it to borrow protected language and writing style just as freely. Building data centers and power plants is much more expensive and time-consuming than hiring lawyers to draft licensing agreements with news organizations. Fair use does not permit this kind of harmful and substitutive copying, retention, and regurgitation of a work-let alone of anything humanity has ever produced. In competing with China, the United States weakens itself by abandoning the intellectual property protections that fuel innovation and sustain American creative companies.
The combined valuation of the top six AI companies is $11 trillion - more than three times France's GDP. Private investment in AI in the United States reached nearly $350 billion in 2025 and is accelerating in 2026. Therefore, intellectual property theft certainly does not occur for lack of money to pay for it. While licensing deals with publishers are not public, based on the size of the few deals that have been disclosed, it is estimated that less than half a percent of that investment is going to compensate the people and companies that create the data that powers AI.
While there are many sources of data, AI executives themselves have recognized that high-quality, original content is particularly valuable to the effectiveness and reliability of the technology. Five of the top ten sites used to train some of the most popular language models belong to news publishers. OpenAI confessed that it would be "impossible to train today's leading AI models without using copyrighted materials." An engineer at the company wrote that the models' success "is not determined by architecture, hyperparameters, or optimization choices. It is determined by your data set, nothing more." In other words: you are what you eat.
The case of 'The New York Times'
Let's look closely at the experience of "The New York Times" to understand how this works.
If you want comprehensive, accurate answers in your AI chatbot, it's hard to imagine a better data source than a news organization that, for 175 years, has employed seasoned, well-paid professional journalists to uncover new information, chronicle ongoing events, and assess developments in politics, business, culture, sports, science and global affairs. This original work is valuable to technology companies in large part because it is carefully written and edited, independently verified, held to the highest standards of fairness and accuracy, and presented in a distinctive and engaging way.
Last year alone, The New York Times published nearly half a million of these works - from articles to photos, videos and podcasts - at a cost of more than $2 billion. We have journalists in all 50 American states and 155 countries, and these professionals often face life-threatening situations. In Ukraine, for example, we had more than 70 journalists and support staff in the field. All this in 2025 alone. Add up these contributions over 175 years and 20 million original works, and you have a clearer idea of ​​what our newsroom has contributed to public understanding of the world.
The distinctive value of Times journalism - like that of other sources of quality journalism - has been repeatedly reaffirmed by the preference that AI companies show for it.

While most AI companies hide their training sources, "The Times" was the largest single source of proprietary data in a major dataset used to train several different models, followed by other news organizations such as "The Guardian" and "Los Angeles Times." AI companies consider extracting information from quality news organizations to be one of the most reliable signs that their products are working properly. As a Microsoft vice president put it: "Premium content significantly improves the quality of responses."
However, tech giants have consistently argued that they should not be required to ask for permission to use - much less pay for - this type of intellectual property. Their argument, as their actions show, is that they are entitled to it. Meta trained its model on a notorious database of illegally pirated books. Perplexity openly challenged the established norm that websites cannot be crawled surreptitiously, contrary to its explicit objections. OpenAI lobbied the US government to obtain legal immunity for confiscating other people's works. Even Anthropic, often cited for its commitment to the ethical development of AI, has refused to pay for the high-quality journalism it uses in its products.
Actions like these led the Times to sue OpenAI, its partner Microsoft, and later Perplexity, for egregious violations of our intellectual property rights protected by U.S. copyright law - both in training their models and in the continued use of our work in their products. Like other news organizations that have filed similar lawsuits, we believe these breaches threaten the long-term ability of news organizations to continue producing the original, trustworthy journalism that the public - and, as it turns out, AI models themselves - depend on. But legal proceedings are slow and expensive - ours has now spanned two and a half years and cost more than $20 million. As AI companies surely know, most news organizations don't have the resources to go to court to defend their rights.
An already fragile sector
Even before the arrival of AI, the global news industry was struggling to survive waves of change unleashed by the internet, the smartphone and social media.
Over the past two decades, the United States has lost, by some estimates, 75% of its journalists and more than 3,000 newspapers. A new newspaper closes every three days. Digital vehicles have not filled even a fraction of this void. Large swaths of the United States no longer have a single reporter asking questions at city hall, covering local schools, or connecting their community with a common set of facts. And when you look at the most expensive and challenging forms of journalism - investigating wrongdoing or going to the front lines of conflict - you see that the number of journalists doing this work has fallen even more dramatically.
The disruption brought about by AI promises to be even more devastating. Before AI, there was a real - if unbalanced - exchange of value between technology platforms and digital content creators like news organizations. This was the pact of the so-called open web. Tech companies - particularly search and social media platforms - took a growing share of advertising revenue that previously went to news organizations, but in return, delivered a much larger audience.
In the next phase of disruption, technology companies, by taking ownership of journalism itself, are also taking a growing share of the audience it captures.
Take the case of Google. The goal of search engines has always been to identify the most useful sites and send people to them.

People would go to Google, search for a topic and click on a link to sites like the "Financial Times", "Le Monde" or "El País" to read the article. Google kept the vast majority of advertising revenue. But it also sent significant traffic to news organizations through links, allowing publishers to make money by displaying ads or selling subscriptions.
In the age of AI, Google increasingly uses content from news organizations and other websites to answer questions directly. As a result, getting a Google user to click on a link is, according to industry research, ten times more difficult today than it was a decade ago. Still, Google holds itself to the highest standard in terms of sending readers to publishers, and we can only hope that commitment continues. Competing AI models send referral traffic at a rate 96% lower than Google search, according to one study.
Tech giants are keenly aware of the implications of this shift on news organizations' already fragile business models. As Microsoft's head of AI monetization wrote: "The open web was built on an implicit value exchange, where publishers made content accessible and distribution channels - like search - helped people find it. That model doesn't translate cleanly to an AI-driven world." He added: "Publishers need sustainable and transparent ways to control how their premium content is used." A worthy feeling. But just look at a recent launch page for Microsoft's own AI search engine to find a very different stance: "Hello from Bing! Instead of clicking on links, we can talk about anything you want to know."
This dynamic has evidently caused traffic to news sites to plummet. The largest newspapers tracked by Comscore have seen declines of more than 45% on average as the AI ​​race has intensified over the past four years. Global news publishers polled by the Reuters Institute are bracing for significant traffic declines to continue in the coming years.
Less traffic for publishers likely means fewer opportunities for advertising, which remains an important source of revenue for most news organizations. Over the past two decades, newspapers' combined advertising revenue has fallen by 80%. Meta alone earns eight times more in advertising revenue than all the newspapers in the world combined.
To offset the decline in advertising, many news organizations have turned to subscription models. But as people realize they can access stolen works for free through AI products, it will become increasingly difficult for news organizations to develop and deepen relationships with potential subscribers. This theft doesn't just happen because publishers "leave their toys in the yard"; it happens even when they are "safely locked indoors." One study found that about 30% of scans by AI bots violate explicit restrictions on access to website content, including content protected by paywalls.
The source of revenue that some hope to offset these losses is money from AI companies themselves, through content licensing or micropayments. Some larger news organizations, including the Times, have signed licensing deals. Others have adopted micropayments from AI companies for each individual use of journalism. But there are good reasons to question whether any of these models will be enough to offset the revenue and readership lost to competing AI products.

Meanwhile, many smaller news organizations whose work has also been taken and used by AI models have received no compensation, and the vast majority of publishers say they do not expect significant revenue from AI platforms.
Worryingly, even as these tech companies try to publicize deals and other actions that signal they value journalism, they are simultaneously arguing in court, to lawmakers and federal agencies, that they have no obligation to the creators of the intellectual property they use to power their products.
It's not competition - it's free riding
For To be clear: I'm not raising these concerns because news organizations should fear competition. If tech companies were committing real resources to putting their own reporters in the field to produce original journalism, I would welcome that. But that's not what's happening. Tech platforms have never made serious attempts to create the original, grassroots work - like local reporting, investigative journalism or rigorous product testing - that their users, platforms and AI products depend on. And now they go a step further, simply taking the reporting and coverage of others, often even presenting it as their own. One study found that OpenAI credited the news organizations that unearthed the cited information in just 1% of its responses.
Leaders of previous technology transitions at least tried to argue that their platforms would be symbiotic with creators. Spotify, for example - which has its critics in the music industry - highlights the payments it sends to artists. AI companies, in contrast, have adopted a more openly parasitic stance, closer to that of Napster, the ancient pirate music platform. A senior Microsoft researcher wrote that one of the "central promises of LLMs" is their ability to use "their training data to replace the paid labor of those who created that data." More evocatively, science fiction writer Margaret Atwood likened this dynamic to being "murdered by my replica."
It's a safe bet that such actions by tech giants will fuel destructive trends that are already straining society. A continued decline in original journalism. A rising tide of misinformation, propaganda, conspiracy theories, deepfakes and computer-generated garbage. A public that continues to be radicalized by algorithms that amplify fear, anger and division.
Reporters are responsible for enriching the public record with previously unknown information. That surprising fact. That revealing detail. That eyewitness quote. That secret document. That expert analysis. That photo, video, audio recording. Simply put, original journalism is often how you know what you know. AI products cannot do this type of original reporting. They extract the public record, but have difficulty adding anything to it.
Even the extraction has been problematic. A survey by the "European Broadcasting Union" found that top AI assistants significantly distorted the news in almost half of all responses. Both Google and Apple, for example, have made serious mistakes when using AI tools to rewrite headlines and news alerts from news organizations that appear in their products. Because AI tends to be bad at expressing uncertainty, it's often not just wrong - it's confidently wrong. And unlike the news organizations they steal from, AI companies do not track or correct these errors, leaving their users without any way of knowing when they have been misled.
This matters in part because AI products will likely not just supplement but replace direct relationships with news organizations for many people.

Research suggests this shift is happening much faster than most realize.
Amazon Web Services, which works with many AI companies, estimates that the majority of online content is already generated by AI - a number that some experts expect will reach more than 90% in the coming years. Already today, the number of fake local news sites is greater than the number of real sites, as AI makes it harder for real sites to survive and makes it easier for fake sites to be created at low cost.
Revealingly, AI companies do not want to say that the results of their products are trustworthy. They do not mean they are fair or accurate. This is partly because they are not. When American political activist Charlie Kirk was murdered last year, for example, Perplexity's bot suggested that the White House's statement about Kirk's death had been fabricated, and X's Grok insisted he was alive and well. But just as importantly, AI companies refuse to be responsible for what their chatbots say to users in an attempt to escape legal liability. Microsoft warned when releasing Copilot: "For entertainment purposes only. It may make mistakes and may not work as intended. Do not rely on Copilot for important advice. Use Copilot at your own risk."
At some level, the public understands that this will not be good for them. Two-thirds of Americans are very concerned about AI spreading inaccurate information, according to the Pew Research Center. But a growing percentage of people turn to AI for news, information and guidance - and some consider it more trustworthy than the news organizations it relies on for answers. All of this will worsen the alarming decline in social and civic health. Evidence shows that when a local news organization disappears, people in a community begin to trust each other less and hate each other more. They become more isolated and less tolerant. Civic engagement declines and public corruption increases.
And imagine what happens when technology companies' approach to the journalism industry reaches its logical conclusion. Despite the importance of journalism to the world's most valuable technology, the actions of technology companies are compromising their most important source of new news, new information, new analysis. This would make AI products themselves less useful and less reliable - yet another unnecessary victim of unnecessary and harmful choices.
What we can do
A declining newspaper industry may seem powerless in the face of some of the richest companies the world has ever seen. And the path forward is not made any easier by the reality that we need to continue operating in an information ecosystem controlled disproportionately by these technology giants. But there are still actions we can take - both to stand up against the abuses of AI companies and to prepare our own organizations to succeed in this new era. I will share some ideas on each of these fronts, with the conviction that better and more ideas will emerge from the people in this room.
When it comes to defending your work from technology companies, I have four central thoughts:
Defend your rights. Intellectual property rights need to be maintained if our profession is to have a path forward. In my country, these rights are anchored in the Constitution and supported by centuries of precedent. They are also compatible with a basic ethical understanding that stealing is wrong. But your rights will only be maintained if you insist that they are respected and resist when they are not. This will take courage - and sometimes resources, which are in short supply - but the alternative path of silently tolerating the systematic theft of your work will ultimately undermine your ability to continue doing journalism.
Negotiate carefully.

News organizations that sign deals to license content to AI companies are doing something reasonable. But I advise you to evaluate the long-term viability of each agreement. Tech giants have an extraordinary position of strength: they have already taken your content and intend to use it anyway. Still, before accepting an offer, it's worth asking whether the payment reflects anything close to fair value-and whether you're retaining any meaningful control over how your work will be used.
Press your lawmakers. AI is increasingly unpopular with the public. As lawmakers consider how to respond, our industry needs to unite around a small, clear set of orders. Some initial ideas: Ensure that already robust intellectual property protections are strengthened-not weakened-for the AI ​​era. Require bots to identify themselves and limit their ability to scour websites without permission. Demand transparency so news organizations know when and how their work is used by AI. Ensure AI companies are legally responsible for the defamatory content they generate.
Join others. We face AI companies that spend unimaginable amounts on marketing, lobbying, and political donations to persuade the public and co-opt politicians. The venture capital firm behind many AI investments is now the largest political donor in the United States. The journalism industry's only way to counterbalance this influence is to work together and, equally important, with other creative industries. Participate in amicus curiae briefs and be active in your professional associations. Study how our colleagues in music and other professions navigated their "Napster" moments.
There are also things we can do to make our own news organizations more resilient as we face this challenge. Four more ideas:
Use AI the right way. Newsrooms must create careful standards for the responsible use of AI. And then they must be aggressive and creative to put technology at the service of improving their journalism and strengthening their business. AI can bring real value to organizations that find the right ways to adopt it, and a change of this magnitude will destroy any organization that refuses to evolve. There is nothing inherently bad about AI technology - it is the actions of the companies behind it that need to be reformed.
Be a destination, first and foremost. A world increasingly brokered by AI platforms would leave news organizations even more at the mercy of tech giants to share traffic, credit and money. The clearest way to sustain quality journalism will be through direct relationships with the public. Being a destination doesn't mean ignoring the broader internet. You still need to create new relationships where people are, which is usually a technology platform. But to deepen these relationships - make them loyal, habitual and valuable - your audience needs to learn that it's better to engage with you directly rather than through an intermediary.
Focus on original journalism. Many news organizations have weakened themselves and become commodities by trying to feed the ever-changing preferences of search algorithms and social networks with clickbait, aggregation and easy opinions. The economics of this approach will get even worse. To be a destination in an AI-mediated world, you will need journalism so distinctive that it has its own gravity. The heart of it is original journalism. The public has no other source for this work. And neither does AI.
Explain why journalism matters. AI companies have gigantic megaphones and have been very carefully - and selectively - communicating the benefits of their work while minimizing the harms.

The news industry must, in turn, show that original journalism is an essential ingredient in healthy societies, safe nations and strong democracies - and demonstrate how the actions of tech giants are putting all of this at risk.
Information is valuable. Journalism is valuable
In the latest digital transition, news organizations - including the Times, for a long time - have bought into Silicon Valley's oft-repeated assertion that "information wants to be free." Many didn't even know that the original quote, from technology philosopher Stewart Brand, had another part: "Information wants to be expensive, because it's so valuable - the right information in the right place simply transforms your life."
We can't be so naive this time. News organizations are collectively smaller and weaker than they were two decades ago. Tech giants are bigger and stronger - and much more willing to use their size and power. Meanwhile, the AI ​​wave itself could be bigger and faster as the technology continues to improve. Even if things seem fine for now, remember: these first waves herald an approaching tsunami.
As we prepare, we need to remember: information is valuable. Journalism is valuable.
The internet is already overloaded with bots and digital trash. It is increasingly difficult to know where things came from and whether they are true. This has created a growing sense that nothing can be trusted, requiring everyone to have an almost paranoid vigilance over everything - or, worse, a descent into nihilism. The effect is not just that people believe false things: it is that they stop believing true things. This toxic combination is already leading more people to completely disengage. Tech companies wave to these trends and say "it's not our fault" and, even more tellingly, "it's not our problem."
News organizations should position themselves as the credible alternative in this chaos. News and information you can trust is rarer and more necessary than ever. The kind produced by teams of experienced professionals, backed by rigorous processes and standards. According to research, when someone wants to verify something they've found that they think might be fake, the preferred option is "a news source I trust." Last on the list? An AI chatbot.
I remain convinced of the value created by quality news organizations dedicated to the difficult and expensive work of original journalism - for readers, for communities, for society as a whole. And, yes, even for AI models.
Who else will go to the places where events are unfolding? Who will bring us first-hand accounts from the front lines of a war? Who will equip us with reliable information in a public health crisis? Who will expose the successful company or political career built on a lie? Who will ensure that economic policy debates are informed by their impacts on real people? Who else can enrich all this work with hard-won expertise that adds perspective and context, and with deeply rooted professional commitments to making every story as fair and accurate as possible?
The question is whether that value will be sucked away by the tech giants - or will go back to news organizations, allowing them to continue this essential work.
I hope you all take this question seriously. I believe the future of our news organizations and the health of the public sphere depend on how we respond. Thank you.
(c) 2026 The New York Times Company.
Original text available at: https://www.nytco.com/press/a-i-journalism-and-the-uncertain-future-of-the-public-square/

Source: G1

Continue lendo

03/06/2026

UFOs land in the sea in Paraná in another record after a case in a rural area

A video recorded in the early hours of last Sunday (31) recorded unusual lights over the sea in Pontal do Paraná (PR)....

03/06/2026

Brazilian Nickel rejects barrier to Asian nickel and advocates free trade

Amid the debate on trade defense measures against Asian products, Brazilian Nickel defends that Brazil does not adopt...

03/06/2026

Sthefany Brito announces the birth of her third child

Actress Sthefany Brito, 38, announced the birth of her third child this Wednesday (3). An Instagram post reveals that...

Top Cifras

Toque agora.

1 The King Of Broken Hearts George Strait

2 Desenganos Da Cor Do Pagode

3 Gasoline Halsey

4 Salmo 100 Catalau e Zeider